Eclipse Community Forums: EMF » [CDO] Unmonitored fail-over example

Home » Modeling » EMF » [CDO] Unmonitored fail-over example

[CDO] Unmonitored fail-over example [message #954950]

Tue, 23 October 2012 07:38

Eclipse User

Hi,

I just tried to run the unmonitored fail-over example in org.eclipse.cdo.examples.server.FailoverExample (InitialMaster, InitialBackup).

First, I run the InitalMaster main program. I get the following exception

Exception in thread "main" org.eclipse.net4j.util.lifecycle.LifecycleException: Could not activate TCPClientConnector[localhost:2.037]

. That means the InitialMaster-Repo can not connect to the InitialBackup-Repo. This is because the InitialBackup main program is not running yet and has not created an acceptor. It is the same vice versa.

So, how can I run this example?

Cheers,
Tino

[Updated on: Fri, 26 October 2012 03:56] by Moderator

Re: [CDO] Unmonitored fail-over example [message #956224 is a reply to message #954950]

Wed, 24 October 2012 06:33

Eclipse User

Hi Tino,

I'm currently off at the EclipseCon and can not do deep code investigations. I'll follow up on this issue next week...

Cheers
/Eike

----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper

Am 23.10.2012 13:38, schrieb Tino Milschewski:
> Hi,
>
> I just tried to run the unmonitored fail-over example in org.eclipse.cdo.examples.server.FailoverExample
> (InitialMaster, InitialBackup). First, I run the InitalMaster main program. But, I get the following exception
> Exception in thread "main" org.eclipse.net4j.util.lifecycle.LifecycleException: Could not activate
> TCPClientConnector[localhost:2.037]. I think this is because the InitialBackup main program is not running yet and has
> not created an acceptor. It is the same vice versa.
>
> How can I run this example?
>
> Cheers,
> Tino

Re: [CDO] Unmonitored fail-over example [message #977857 is a reply to message #956224]

Fri, 09 November 2012 11:37

Eclipse User

Hi Eike,

I am using the monitored fail-over example now and it works fine so far. But I recognized an issue. I perform the following steps:

1.) Deleting both repos in the '_database' folder
2.) Starting the monitor
3.) Starting agent1 and committing a customer
4.) Starting agent2

Then agent2 shows the following errors:

Connecting to network...
Type changed to BACKUP
Connected
State changed to OFFLINE
[INFO] Connected to master.
State changed to SYNCING
[WARN] Replication attempt failed. Retrying in 3 seconds...
org.eclipse.net4j.db.DBException: org.h2.jdbc.JdbcSQLException: Tabelle SUPPLIER nicht gefunden
Table SUPPLIER not found; SQL statement:

Cheers,
Tino

Re: [CDO] Unmonitored fail-over example [message #977916 is a reply to message #977857]

Fri, 09 November 2012 12:34

Eclipse User

Hi Tino,

This week I've been working on the offline/failover modes and I've fixed quite some issues. Most of them are already
committed to 4.2, there's one thing I'm still working on. Maybe your specific problem is already solved. When I'm done
I'll publish the new RCP examples applications for offline/failover. Please bear with me...

Cheers
/Eike

----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper

Am 09.11.2012 17:37, schrieb Tino Milschewski:
> Hi Eike,
>
> I am using the monitored fail-over example now and it works fine so far. But I recognized an issue. I perform the
> following steps:
>
> 1.) Deleting both repos in the '_database' folder
> 2.) Starting the monitor
> 3.) Starting agent1 and committing a customer
> 4.) Starting agent2
>
> Then agent2 shows the following errors:
> Connecting to network...
> Type changed to BACKUP
> Connected
> State changed to OFFLINE
> [INFO] Connected to master.
> State changed to SYNCING
> [WARN] Replication attempt failed. Retrying in 3 seconds...
> org.eclipse.net4j.db.DBException: org.h2.jdbc.JdbcSQLException: Tabelle SUPPLIER nicht gefunden
> Table SUPPLIER not found; SQL statement:
>
> Cheers,
> Tino

Re: [CDO] Unmonitored fail-over example [message #986294 is a reply to message #977916]

Mon, 19 November 2012 13:34

Eclipse User

Am 09.11.2012 18:34, schrieb Eike Stepper:
> Hi Tino,
>
> This week I've been working on the offline/failover modes and I've fixed quite some issues. Most of them are already
> committed to 4.2, there's one thing I'm still working on. Maybe your specific problem is already solved. When I'm done
> I'll publish the new RCP examples applications for offline/failover. Please bear with me...
I published the new example and described it here: http://thegordian.blogspot.de/2012/11/use-online-data-offline.html

An additional demo for the fail-over setups is at http://www.youtube.com/watch?v=2XSFiPlJIKs

Cheers
/Eike

----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper

Re: [CDO] Unmonitored fail-over example [message #986443 is a reply to message #986294]

Tue, 20 November 2012 07:50

Eclipse User

Hi Eike,

thanks. Everything seems to work so far Smile

.

Cheers,
Tino

Re: [CDO] Unmonitored fail-over example [message #986446 is a reply to message #986443]

Tue, 20 November 2012 07:53

Eclipse User

Am 20.11.2012 13:50, schrieb Tino Milschewski:
> Hi Eike,
>
> thanks. Everything seems to work so far :).
Excellent!

Cheers
/Eike

----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper

Re: [CDO] Unmonitored fail-over example [message #990221 is a reply to message #986443]

Tue, 11 December 2012 08:28

Eclipse User

Hi Eike,

I am still using the monitored fail-over example (4.2-M3 (S20121114-1534)).

I modified the client code: Now the client initially commits 1000 new customers. The database files of the master repository grow in size as expected (to more than 100 MB). But the database files of the backup repository remain at the initial size of 500 KB. Nevertheless, the backup repository seems to cache the data, because it is still available when I shut down the master repo.

Is this the expected behaviour or am I missing something?

Edit: I think the client session caches the objects. So, maybe the synching does not work anymore?

Cheers,
Tino

[Updated on: Tue, 11 December 2012 09:29] by Moderator

Re: [CDO] Unmonitored fail-over example [message #990290 is a reply to message #990221]

Tue, 11 December 2012 13:01

Eclipse User

Am 11.12.2012 14:28, schrieb Tino Milschewski:
> Hi Eike,
>
> I am still using the monitored fail-over example (4.2-M3 (S20121114-1534)).
>
> I modified the client code: Now the client initially commits 1000 new customers. The database files of the master
> repository grow in size as expected (to more than 100 MB). But the database files of the backup repository remain at
> the initial size of 500 KB. Nevertheless, the backup repository seems to cache the data, because it is still available
> when I shut down the master repo.
>
> Is this the expected behaviour or am I missing something?
I have no filesystem expectations whatsoever ;-)

Hard to predict what H2 does and how. 500KB sounds very little, though. Can you restart the repo to ensure that it
doesn't just use its in-memory caches?

Cheers
/Eike

----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper

Re: [CDO] Unmonitored fail-over example [message #990848 is a reply to message #990290]

Fri, 14 December 2012 07:54

Eclipse User

Hi Eike,

when I shutdown everything and only restart the monitor and the former backup repo, there is no data in the repo. I am pretty sure that the repos are not synchronizing. And there is no "State changed to SYNCING" in the log as above.

I switched my target platform back to 4.2-M2 (S20121002-0801). Here, the backup repo grows in size and it says "SYNCING".

Cheers,
Tino

Re: [CDO] Unmonitored fail-over example [message #990938 is a reply to message #990848]

Sat, 15 December 2012 01:54

Eclipse User

Am 14.12.2012 13:54, schrieb Tino Milschewski:
> Hi Eike,
>
> when I shutdown everything and only restart the monitor and the former backup repo, there is no data in the repo. I am
> pretty sure that the repos are not synchronizing. And there is no "State changed to SYNCING" in the log as above.
>
> I switched my target platform back to 4.2-M2 (S20121002-0801). Here, the backup repo grows in size and it says "SYNCING".
I get the sense that you're trying out the "old" fail-over example. I didn't verify if that still works correctly. You
should instead play with the new offline and fail-over example application that I announced on my blog.

Cheers
/Eike

----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper

Re: [CDO] Unmonitored fail-over example [message #992432 is a reply to message #990938]

Fri, 21 December 2012 11:09

Eclipse User

Hi Eike,

I checked the org.eclipse.emf.cdo.examples.client.offline example with one monitor, two fail-over agents and one client. I modified the client code to commit 1000 customers. The backup fail-over repo still does not grow while the master fail-over grows to more than 200 MB. And the CDO Server Browser also shows that there are no customers in the backup repo.

Cheers,
Tino

Re: [CDO] Unmonitored fail-over example [message #992488 is a reply to message #992432]

Fri, 21 December 2012 14:25

Eclipse User

Am 21.12.2012 17:09, schrieb Tino Milschewski:
> Hi Eike,
>
> I checked the org.eclipse.emf.cdo.examples.client.offline example with one monitor, two fail-over agents and one
> client. I modified the client code to commit 1000 customers. The backup fail-over repo still does not grow while the
> master fail-over grows to more than 200 MB. And the CDO Server Browser also shows that there are no customers in the
> backup repo.
I've double-checked this and I cannot reproduce it. All involved nodes behave as expected and the server browsers give
the same evidence. Without more infos I can hardly help. Maybe you have the chance to debug it and find the cause.

Cheers
/Eike

----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper

Re: [CDO] Unmonitored fail-over example [message #1005473 is a reply to message #992488]

Fri, 25 January 2013 10:27

Eclipse User

Hi Eike,

currently we are using the bundles directly from the git-repository, instead of 4.2-M3 (S20121002-0801). Now, it works in the example and it also seems to work in our application Smile

.

By the way: In our scenario, it could happen that the monitor fails. After restarting the monitor, it would be nice, if the fail-over repos could automatically reconnect to it. What do you think?

Cheers,
Tino

Re: [CDO] Unmonitored fail-over example [message #1005485 is a reply to message #1005473]

Fri, 25 January 2013 10:57

Eclipse User

Am 25.01.2013 16:27, schrieb Tino Milschewski:
> Hi Eike,
>
> currently we are using the bundles directly from the git-repository, instead of 4.2-M3 (S20121002-0801). Now, it works
> in the example and it also seems to work in our application :).
Excellent ;-)

> By the way: In our scenario, it could happen that the monitor fails. After restarting the monitor, it would be nice,
> if the fail-over repos could automatically reconnect to it. What do you think?
Sounds reasonable. If that's currently not the case, please submit an enhancement request. Please note that, without
someone contributing a patch, I see no chance to work on this anytime soon.

Cheers
/Eike

----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper

Re: [CDO] Unmonitored fail-over example [message #1005721 is a reply to message #1005485]

Mon, 28 January 2013 06:54

Eclipse User

I have opened an enhancement request: https://bugs.eclipse.org/bugs/show_bug.cgi?id=399241

Cheers,
Tino

Previous Topic:	Mapping eType of reference to eType of attribute
Next Topic:	Modifying the generated XML (model) from diagram

Goto Forum:

-=] Back to Top [=-

Current Time: Wed Jul 23 12:10:58 EDT 2025

.:: Contact :: Home ::.

Breadcrumbs

Sign up to our Newsletter